Skip to main content
Internet Archive's 25th Anniversary Logo

Full text of "A faint red stellar halo around an edge-on disc galaxy in the Hubble Ultra Deep Field"

See other formats


Mob. Not. R. Astron. Soc. 000, dEl (2004) Printed 2 February 2008 (MN lAT^X style file v2.2) 

A faint red stellar halo around an edge-on disc galaxy in 
the Hubble Ultra Deep Field 



o 
o 

(N 

c 

00 



> 

o 

(N 

\o 
o 
^1- 

o 

Oh 

6 



X 



Stefano Zibetti 1 * and Annette M. N. Ferguson 1 

1 Max-Planck-Institut fur Astrophysik, Karl-Schwarzs child- Str. 1, D-85748 Garching bei Miinchen, Germany 



Accepted . Received ; in original form 2004 April 23 



ABSTRACT 

We analyse the detailed structure of a highly- inclined (i > 80°) disc galaxy which 
lies within the Hubble Ultra Deep Field (UDF). The unprecedented depth of the UDF 
data allow disc and extraplanar emission to be reliably traced to surface brightness 
levels of fiv.i,z ~ 29 — 30 mag arcsec -2 (corresponding to rest-frame equivalents of 
Hg,r,i ~ 28 — 29 mag arcsec -2 ) in this redshift z — 0.32 system. We detect excess 
emission above the disc which is characterised by a moderately-flattened (b/a ~ 0.6) 
power-law (I oc R~ 26 ). The structure and colour of this component are very sim- 
ilar to the stellar halo detected in an SDSS stacking analysis of local disc galaxies 
(Zibetti. White. & Brinkmann 2004) and lend support to the idea that we have de- 
tected a stellar halo in this distant system. Although the peculiar colours of the halo 
are difficult to understand in terms of normal stellar populations, the consistency found 
between the UDF and SDSS analyses suggests that they cannot be easily discounted. 



Key words: 

spiral 



galaxies: haloes, galaxies: structure, galaxies: photometry, galaxies: 



1 INTRODUCTION 

In the currently-favoured ACDM framework, galaxy forma- 
tion proceeds hierarchically with small structures forming 
first and later merging and accreting to form large galax- 
ies. In these models, a significant fraction the stars which 
reside in the stellar halo and thick disc are tidally-stripped 
from small satellites as they fall within the h ost potential 
(e.g. lAbadi. Navarro. Steinmetz fc EkdEoOal . The struc- 
ture, composition and ubiquity of these faint stellar compo- 
nents are thus expecte d to reflect the details of t he galaxy 
assembly process (e.g. iBullock fc Johnstonl l2004'l . 

Observations of stellar haloes and thick discs in external 
galaxies are extremely challenging. These faint components 
typically have surface brightnesses fj,n > 28 mag arcsec -2 , 
which corresponds to > 7 magnitudes fainter than the 
sky. Measurements of diffuse emission require flat-fielding 
and sky subtraction uncertainties, as well as scattered 
light effects, to be significantly less than 0.1 per cent. 
Until recently, few observational constraints were avail- 
able with which to confront models. While there have 
been some detections of faint stellar componen ts in exter- 
nal galaxies (e.g. |S ackgtt_et_aL|ll994; Morrison et alJll997t 
iLeaueux et al.lll998r |Ahg_et_alJll999|), the re have also bee n 
non-detections (e.g. IZheng et alJ ll99H iFrv et"afl Il999tl . 
leading one to question to what extent the results re- 



E-mail: zibctti@MPA-Garching.MPG.DE 



fleet real variance as opposed to systematic effects (see 
IZibetti. White, fc Brinkmannl20ol hereafter Z04, for a full 
discussion). The alternative technique of using wide-area re- 
solved star counts to study stellar haloes sidesteps the diffi- 
culties inherent in quantifying extremely faint diffuse emis- 
sion, but can only be appli ed to a handful of nearby galaxies 
(e.g. iFereuson et"ai]|2002t . 

Recently Z04 have conducted the first statistical study 
of stellar halo emission by stacking > 1000 homogeneously 
rescaled edge-on galaxies from the Sloan Digital Sky Sur- 
vey (SDSS). This technique has allowed quantitative anal- 
ysis of the "mean" stellar halo to /i r ~ 31 mag arcsec -2 . 
These authors find a halo characterised by a moderately- 
flattened (b/a ~ 0.6) power-law (/ oc R~ 2 ) and puzzling 
colours (g — r — 0.65, r — i = 0.6) that cannot be easily 
explained by normal stellar populations. The detection of 
flatter red stellar e nvelopes around extrem e late- type edge- 
on disc galaxies bv lDalcanton fc Bernst ein (2002) may rep- 
resent the higher surface brightness regions of these haloes. 

The recent release of the Hubble Ultra Deep Field 
(UDF), a public survey carried out with the Advanced Cam- 
era for Surveys using Director's Discretionary time in Cy- 
cle 12 (IBeckwith. Somerville fc Stiavellil 1200511 . provides a 
rare opportunity to quantitatively study galaxy structure 
to the depths where halo emission should prevail. This Let- 
ter reports the detection of a faint red stellar halo around a 
z — 0.32 edge-on galaxy in the UDF. 



2 S. Zibetti and A. M. N. Ferguson 



2 OBSERVATIONS AND ANALYSIS 

The Hubble Ultra Deep Field (UDF) consists of 400 orbits 
of integration on a single 11.3 arcmin 2 field lying within the 
Chandra Deep Field South GOODS area, centred at RA= 
03' l 32 m 39 , !0, Dec= -27°47'29'.'l (J2000). Four filters have 
been utilised: B(F435W), U(F606W) (each for 56 orbits, 
37.5 hours) and i(F775W), z(F850LP) (each for 144 orbits, 
96.4 hours). The UDF represents the deepest observations 
yet obtained with HST, or any ground-based telescope. 

We focus on a relatively large, well- formed disc galaxy 
located at RA= 03 h 32 m 4m, Dec= -27°48'52'.'9 (J2000). 
This galaxy is clearly detected in all 4 bands and is iden- 
tified with source #31611 at z = 0.322 in t h e CO MBO- 
17 photometric redshift survey of IWolf et all l|2004f) . The 
high inclination of this system makes it an ideal target for a 
study of faint extraplanar emission. At this redshift, 1 arc- 
sec=4.7 kpc and the lookback time is 3.6 Gyr 1 . We note the 
presence of a ring-like structure of blue knots lying along 
the western major axis (see Fig. ^a)); t his structure has a 
photometric redshift of unity according to lWolf et al.l <|2004T) 
and thus appears as a chance alignment. 

Our analysis is based on the reduced UDF data vl.O 
released to the community by STScI on March 9, 2004 2 . 
The data were pre-processed using the standard ACS/WFC 
pipeline and subsequently combined using the Multidrizzle 
package. The observations were obtained using a 4-point 
dither pattern to provide sub-pixel sampling, as well as a 
larger-scale pattern to cover the 2 arcsec gap between the 
ACS/WFC chips. The final pixel scale after drizzling is 0.6 of 
the original, corresponding to 0.03 arcsec. Note that all mag- 
nitudes are on the AB-system with zero-points provided by 
STScI. We have additionally applie d a co rrection for Galac- 
tic extinction using Schl egel et al.l (1998). 

Using the geometric parameters obtained from 
STSDAS-ellipse fitting of the i-band isophotes, we have ex- 
tracted 1200 x 1200 pixel 2 frames in each band, centred on 
the galaxy and rotated according to the average position 
angle of the isophotes at /i; ~ 24 mag arcsec -2 . The RGB 
colour composite image of the galaxy is shown Fig. (a), 
where we use the B-band image for the blue channel, V for 
the green, and i + z for the red. 

We use SExtractor v2.3 feertin fc Arnoutsll99'r3) to ex- 
tract very low signal-to-noise segmentation images of the 
frames, adopting a 5 x 5 pix 2 gaussian smoothing kernel 
(cr = 2 pixels) , a \-a detection threshold and a 25 pixel min- 
imum detection area. After excluding the segment (s) corre- 
sponding to the galaxy and manually editing the mask to 
include sources that were not correctly deblended from the 
galaxy, we combine the masks in the four bands and "grow" 
the resulting mask by 2 pixels, in order to have better cover- 
age of the extended diffuse sources. The masked i + z-band 
image is shown in Fig. Q(b), with an enhanced contrast in 
order to show the lowest intensity levels. 

Visual inspection indicates that the galaxy is close to 
edge-on and this is confirmed by the measured isophotal 
axial ratio of 0.2 at fii ~ 24 mag arcsec -2 . Assuming the 
standard formula for an infinitely thin disc (cos i=b/a) , this 



1 For fl" =70km sec -1 Mpc -1 , n =l, Q A =0.7 

2 Available from http://www.stsci.edu/hst/udf/release 



corresponds to an inclination of 78°. For the more realis- 
tic case of finite thickness, the inclination will be somewhat 
greater than this. Fig. Qa) also suggests a late-type classifi- 
cation for the galaxy based on prominent star-forming spiral 
arms, a significant amount of dust in the inter-arm region 
and a very small bulge, if present at all. In Fig.Qb), one can 
clearly see the presence of red extraplanar emission which 
will be analysed in significant detail in the following section. 

Although the UDF science images are already back- 
ground subtracted, we estimate and subtract the residual av- 
erage local background using an annulus (inner radius= 400 
pixels or 56 kpc, width = 200 pixels or 28 kpc) around the 
galaxy on the masked images . This annulus is chosen to lie 
outside the region where we might reasonably expect faint 
halo emission to be present. The maximum correction corre- 
sponds to a surface brightness of ~ 31 mag arcsec -2 (mea- 
sured in the z-band); comparison to the faintest measured 
z-band surface brightness indicates that the level of uncer- 
tainty due to the background is no more than 20%. Smaller 
background uncertainties are found in all other bands. 

The primary photometric analysis is performed on the 
masked images using an in-house developed C code which 
provides us with the average pixel value as a function of ra- 
dius within a wedge-shaped region with apex located at the 
galaxy centre. As we believe the galaxy is close to edge-on, 
this technique provides an alternative to making simple ma- 
jor and minor axis cuts and is particularly well-suited to the 
study of very faint emission at large radius since relatively 
more pixels are averaged in these parts. Errors are attached 
by taking into account both the count statistics and the typi- 
cal rms background fluctuations occurring on the same scale 
as the aperture. The latter quantity is computed by evalu- 
ating the surface brightness in a number of non-overlapping 
apertures spread throughout the frame and located at least 
400 pixels from the centre of the galaxy. 



3 RESULTS 

In Fig.|5|and|21 we show the surface brightness profiles mea- 
sured along the major and minor axes of the galaxy respec- 
tively. As described in the legend, the different coloured sym- 
bols correspond to the four bands. The asymmetry seen in 
the inner regions of both profiles attests to the presence of 
spiral structure and dust lanes in these parts. We include 
these regions for completeness, but note their interpretation 
is complicated; our main focus is the behaviour of the pro- 
files at very large radius. 

The major axis profile is extracted using wedges aligned 
with the +x and — x-axes and with a small opening angle of 
10° (see Fig. |2J • While a warp is present in the outer parts 
of the disk, our extraction method involves summing over 
many pixels at large radius and is thus not significantly af- 
fected by the deviation from the midplane. In all four bands, 
the profile shows a profound change in slope at ~ 3 arcsec. 
By fitting an exponential function to the surface brightness 
profile of the bright inner disc (corresponding to the region 
where highly structured spiral arms and dust lanes are vis- 
ible in Fig0, we find that the projected scalelength varies 
with wavelength from r- m ner,B = 4.2 arcsec to rinner.z = 1-3 
arcsec. The steeper outer profile is very similar in all bands, 
with exponential scalelength of ~ 0.55 arcsec. Beyond 5 arc- 



- 6 


1 


- 


'- i 




* " n J 


+ 






- 2 




- 


: ° 


^■^^■^ 


- 


■ 
- -2 




1 






> 


'- -S -6* 4 

■. . 1 . . . 1 . . . 1 . 


-2 2 J 

. . 1 . . . 1 . . . 1 . . . 


E^ 

l 6 a - 

.... i -■_ . i .. .- 



Faint red halo around an edge- on disc in the UDF 3 



(.-■■.-.-. — i . ■ ■ . ■ i y - '■-:.-■>.: i, -.»,-;;■■. t-^-y o ■ ■ ■ 







.-^-: 



« 



a 



^ ! . t . . : :1 : • . ■?. ■' 



V '.,£ 4 ■■.'■■fr' v 'p8 



Figure 1. a) Full resolution colour composite RGB ((i + z)-V-B) image of the galaxy showing whole dynamic range, b) i + z band 
inverted gray-scale image, with enhanced contrast in the low surface brightness levels and masked regions superposed. Intensity scaling 
is a square root function in both panels. Axes show the offset in arcseconds with respect to the galaxy centre. 



30 



22 



b 24 



iB8 



2B 



30 



32 



A 



z band (ju-1) 
i band 

V band (/i+l) 
~B band 0»+B). 

w 



I ' ' ' I 



111= 



rV 



-4-2 £ 
R [arcsec] 



6 9 



Figure 2. Major axis surface brightness profiles. Negative R are 
used for the eastern side, positive for the western. Error bars are 
computed as explained in the text. Only points with S/N>1 are 
plotted. The profiles for the different bands are coded in different 
colours and offset in order to avoid confusion, as indicated in the 
legend. 



sec an excess with respect to the exponential profile is seen 
in the three reddest bands at the level of fi ~ 29 — 30 mag 
arcsec -2 . 

The minor axis profile is extracted using wedges aligned 
with the the +y and — j/-axes and with a wider opening 
angle of 45° in order to ensure adequate S/N far from the 
disc plane (see Fig. |3J. On the southern side of the galaxy, 
the inner profile is more regular and declines as a smooth 
exponential until approximately 0.5 arcsec. Beyond 1 arcsec, 
a significant power-law excess is apparent (on both sides 
of the plane) in all four bands (see Fig. [3J. The slope of 
this power-law component is measured to be ~ 2.6 in the 



20 



22 



b 24 



,28 



2B 



30 



32 



i | i | i | r 

z band (ju— 1)_ 
* i band 
-\ V band Gu4i)_ 
S* B band (n+2) 












ii 



rl 



J 






II 



-3 -S -1 1 2 3 

R [arcsec] 

Figure 3. Same as Fig|2]for the minor axis. Negative R are used 
for the southern side, positive for the northern. 



three reddest bands; we expect the effect of dust (as well 
as emission from an inclined disc) to be minimal at these 
heights (> 5 kpc) from the midplane. 

We have checked that the point spread function (PSF) 
does not have a significant effect on our derived profiles by 
studying the surface brightness profile of a relatively bright, 
unsaturated star (mv ~ 20.5 mag) which falls within the 
UDF. We find that the surface brightness drops by > 10(12) 
mag arcsec^ 2 from the centre to 1 (2) arcsec in all bands. 
As a result, scattered light from the central PSF should be 
well below our measurement uncertainties. 

In order to estimate the isophotal ellipticity of the 
power- law component, we measure the surface brightness 
as a function of radius in narrow wedges (13° wide) placed 
at different position angles. By comparing surface bright- 



4 S. Zibetti and A. M. N. Ferguson 




2 3 

R [arcsec] 

Figure 4. Colour profiles: V — i (top), i — z (bottom). Black solid 
lines represent the major axis and red points represent the minor 
axis (filled circles for the northern side, open diamonds for the 
southern). See text for details. 



nesses as a function of position angle out to a radius of 5.5 
arcsec, we derive an axial ratio of b/a ~ 0.45 in the i-band 
at Hi ~ 29. The outermost isophotes thus appear signif- 
icantly rounder than the inner (disc-dominated) isophotes. 
Our measurement represents a lower limit on the actual flat- 
tening however, since low-level disc emission may still con- 
taminate the isophotes at and 180° . Indeed, if we exclude 
the sectors which lie within 25° of the major axis, the ellip- 
ticity b/a of the faint outer component increases to 0.6. 

In Fig. 0]we plot the V — i and i — z colours along the 
major (black solid line) and minor axes (red points) as ob- 
tained from combining the surface brightness profiles shown 
in Figures |2] and The major axis profile represents the 
average of the eastern and western axes in regions where 
S/N> 2. Error bars are attached using the standard propa- 
gation of the errors on the surface brightness measurements. 
The profiles along the major axis display a clear blue gra- 
dient in V — i within ~ 2.5 arcsec, where the bright disc 
dominates. Beyond ~ 3 arcsec, where the exponential slope 
significantly steepens, the gradient in this colour is inverted. 
The i — z profile is significantly flatter, slowly varying from 
0.3-0.35 in the inner 1 arcsec, to 0.15 at 4 arcsec. 
For the minor axis, the north and south sides are shown 
separately as filled circles and open diamonds respectively. 
Along the minor axis we observe a clear blue gradient in V — i 
within ~ 0.5 arcsec, while the i — z profile in this region is 
very irregular and likely reflects both internal structure as 
well as dust within the disc. The transition between the disc 
and power-law regions is characterised by a strong red gra- 
dient in i — z, from 0.4 at 0.5 arcsec to 0.7 at 1 arcsec (con- 
sidering the average of the northern and southern profiles) . 
In the region dominated by the power-law excess (R > 1.5 
arcsec), the colour measurements are rather noisy due to 
the very low surface brightness, but still there is evidence 
for V - I ~ 0.6 - 0.7 and i - z ~ 0.5 ± 0.1. 



We also estimate the total magnitude of the galaxy, 
of the disc alone, and of the galaxy excluding the disc in 
each band (see Table 0. The total magnitude is integrated 
in an ellipse with b/a = 0.6, a = 8 arcsec, while for the 
disc alone we adopt an ellipse with b/a — 0.2, a — 4.8 
arcsec. The "Total— Disc" magnitude is integrated in the 
complementary region. Typical uncertainties are of order of 
0.05 for the "Total— Disc", essentially given by count statis- 
tics and background uncertainty. In the three reddest bands 
the contribution of the non-disc component ranges from 
4.3 to 5.6 per cent of the total light and displays colours 
which are significantly redder than those of the disc, namely 
V - i = 0.73 ± 0.10 vs. 0.64 and i - z = 0.52 ± 0.10 vs. 
0.33. Reddening due to dust is unlikely to affect the colours 
at these radii and, in any case, any realistic distribution of 
dust would lead to bluer colours with increasing distance 
from the disc instead of redder ones. 



4 DISCUSSION 

Our analysis of the UDF galaxy, COMBO-17 #31611, has 
yielded the detection of a very low surface brightness struc- 
tural component in addition to the bright disc. The main ev- 
idence for this component is: i) excess emission with respect 
to an exponential profile detected to 3 arcsec (~14 kpc) 
from the disc plane at the level of /iv,i ~ 29 mag arcsec -2 ; 
ii) the isophotal shape is centrally-concentrated and be- 
comes rounder with increasing radius, reaching b/a ~ 0.6 
at the faintest measured isophote; iii) colours far from the 
plane that are distinct from the disc, i.e. i — z ~ 0.5 and 
V — i ~ 0.7 which are 0.2-0.3 and 0.1 mag redder than the 
disc respectively. Based on these properties, we attribute 
this component to the stellar halo of the galaxy, making it 
the most distant detection of a stellar halo yet known. We 
note the striking resemblance between the halo properties 
derived here and those of the "mean" halo detected in Z04's 
SDSS stacking analysis. As discussed in Z04, these observa- 
tions are also consistent with the few extant observations of 
individual stellar haloes. 

The surface brightness profile of the halo is consistent 
with a power-law falling as / <x R~ 26 . This slope must be 
considered an upper limit since if there is contaminating 
outer disc emission within our aperture, it will steepen the 
observed profiles even beyond 1 arcsec. An R 1 ' law provides 
a poor fit to the data, which display excess emission at the 
faintest surface brightness levels compared to this profile. 
Less concentrated Sersic profiles, which are more appropri- 
ate for the small bulges of late-type galaxies, provide even 
worse fits, thus ruling out the bulge as responsible for the 
measured emission. The fractional contribution of the halo 
to the total galaxy light is around 5 per cent, but this should 
be considered only approximate since we neglect disc con- 
tamination in the region where the halo light is integrated 
and include the inner halo as part of the disc light. Further- 
more, we have no handle on internal extinction which could 
be significant in the inner regions of the galaxy, where the 
disc contribution is calculated. 

In order to conduct a direct comparison between the 
surface brightness and colours of this distant stellar halo 
and those of local galaxies, corrections need to be applied 
for bandpass shifting and surface brightness dimming. Given 



Faint red halo around an edge-on disc in the UDF 5 

Table 1. Integrated photometry and colours: the total and disc fluxes are integrated in elliptical apertures, as described in the text. For 
the "Disc" and "Total— Disc" components, the percentage over the total flux are given as well. 





mag 


V 


% 


mag 


i 

% 


mag 


2 

% 


V -i 
mag 


i — z 
mag 


Total 

Disc 

Total-Disc 


21.14 
21.18 
24.54 




95.7% 
4.3% 


20.50 
20.55 
23.82 


95.4% 
4.6% 


20.16 
20.23 
23.30 


94.4% 

5.6% 


0.64 
0.64 
0.73 


0.34 
0.33 
0.52 



the photometric redshift of 0.322, the four observed bands, 
BViz correspond rather well to the SDSS ugri bands at 
redshift zero. Assuming a 9 Gy r, Z = QAZp, sing l e stel- 
lar population (SSP), we use the lBruzual fc Charlotl f2003) 
(hereafter BC03) models to calculate that observed sur- 
face brightnesses in Viz translate into rest-frame values of 
~ 1.1 mag arcsec -2 brighter in gri. Our measurements of 
fJ-v,i,z ~ 29 — 30 mag arcsec -2 thus correspond to rest-frame 
equivalents of n g , r ,i ~ 28 — 29 mag arcsec -2 . This is precisely 
the surface brightness range in which a power-law halo com- 
ponent starts to dominate the minor axis profile in Z04's 
SDSS stacking analysis and in which direct detections of in- 
dividual haloes have been made in nearby galaxies. 

We proceed to use the BC03 models to interpret the 
colours of the halo emission in terms of stellar populations. 
First, we compare the observed colours with predictions for 
different SSPs with age s of 2-13 Gyr a nd metallicities of 
0.005-2.5 Z Q adopting a lChabrierl J2003h Initial Mass Func- 
tion (IMF). We find none of these models can simultane- 
ously reproduce the intermediate V — i and the very red 
i — z colours. Unless a super-solar metallicity is assumed 
(Z = 2.5Zq), the measured i — z ~ 0.5 is at least 0.15 mag 
redder than the models and thus inconsistent at more than 
the 1.5 a level. Such an anomaly is reminiscent of the find- 
ing of a correspondingly high r — i ~ 0.6 by Z04, despite the 
intermediate g — r ~ 0.65 that could be accounted for by an 
old population with roughly solar metal enrichment. 

Since we have been unable to produce a model that 
provides a good fit to the observed colours at z = 0.32, 
we adopt the following strategy to derive rest-frame colours 
that can be compared to haloes in the local universe. We 
adopt a fiducial reference model of an SSP of age 9 Gyr at 
the redshift of the galaxy, with Z = 0.4 Zq, and a Chabrier 
IMF (BC03); this model minimises the absolute deviations 
from the observed colours. We then evolve the spectrum 
to 2 = and derive the rest-frame colours that would be 
observed through SDSS gri passbands. These colours can 
then be directly compared to the measurements of Z04 for 
the "mean" halo in their SDSS stack. Not unexpectedly, the 
evolved colours do not provide a good match to Z04's obser- 
vations but interestingly the offset between the model and 
observations is of the same magnitude and sense at z = 
as at 2 = 0.32. Specifically, while the model predicts a blue 
colour (V — i or SDSS g — r) that is consistent or just slightly 
redder than the observations, the predicted red colour (i — z 
or SDSS r — i) is too blue by ~ 0.2 mag. 
This kind of colour anomaly, i.e. too much flux in the spec- 
tral region around 7500A with respect to shorter optical 
wavelengths, was also seen in the stella r halo of the nearby 
system NGC 5907 lLeaueux et alJ (Il998h . An attempt to di- 
rectly resolve the population of metal-rich giant stars that 



would give rise to these red colours was unsuccessful, leading 
to the suggestion that the stellar halo in NGC 5907 formed 
with a non-standard IM F that is dominated by M < 0.2Afo 
stars JZepf et al.ll 20001) . While such a truncation could ac- 
count for the lack of halo giant stars, it is unclear whether it 
could also explain the colours observed here, i.e. g — r typical 
of G-K stars yet r — i typical of M stars. Another possibility 
is that the red colour is contaminated by ionised gas emis- 
sion. Deep Ha imagery of near by gala xies often reveal s faint 
extraplanar emission iFerguson. Wvse. fc Gallagher! 1 19961 : 
iMiller fc VeiHeuxll2003i) . however little is known about the 
extent of these ionised "haloes" at very faint flux levels. Fur- 
ther investigation is clearly required in order to understand 
the origin of the peculiar halo colours, which have now been 
measured in three independent studies. 

As a final remark, we note that the surface brightness 
profile of the disc can be fitted by an exponential function 
that changes scalelength abruptly at 3 arcsec ~ 14 kpc 
(corresponding to ~ 2.4 inner scalelengths) . This behaviour 
is consistent with deep imaging studi es of both edge-on 
and face-on discs i n the local universe llKregel et al.l 120021 : 
iPohlen et al.l 1200211 and likely reflects a truncation of the 
bright inner disc. That this radius also represents the point 
where the major axis colour gradient reverses and starts 
to redden is intriguing and suggests that the stellar pop- 
ulations at the far extremity of the disc are of significant age. 

We thank the referee, Michael Pohlen, for useful com- 
ments. The research of AMNF has been supported by a 
Marie Curie Fellowship of the European Community under 
contract number HPMF-CT-2002-01758. 



REFERENCES 

Abadi, M. G., Navarro, J. F., Steinmetz, M., & Eke, V. R. 
2003, ApJ, 597, 21 

Abe F., et al., 1999, AJ, 118, 261 

Beckwith S., Somerville R., Stiavelli M., 2003, STScI 
Newsletter, Vol 20, Issue 04, pi 

Bertin E., Arnouts S., 1996, A&AS, 117, 393 

Bruzual G., Chariot S., 2003, MNRAS, 344, 1000 (BC03) 

Bullock, J. S. & Johnston, K. V. 2004, in proceed- 
ings of "Satellites and Tidal Streams", eds. F. Prada, 
D. Martinez-Delgado & T. Mahoney, |astrc-ph/0401625| 

Chabrier G., 2003, PASP, 115, 763 

Dalcanton J. J., Bernstein R. A., 2002, AJ, 124, 1328 

Ferguson, A. M. N., Wyse, R. F. G., & Gallagher, J. S. 
1996, AJ, 112, 2567 

Ferguson A. M. N., Irwin M. J., Ibata R. A., Lewis G. P., 
Tanvir N. R., 2002, AJ, 124, 1452 



6 S. Zibetti and A. M. N. Ferguson 

Fry A. M., Morrison H. L., Harding P., Boroson T. A., 

1999, AJ, 118, 1209 
Kregel M., van der Kruit P. C, de Grijs R., 2002, MNRAS, 

334, 646 
Lequeux J., Combes F., Dantel-Fort M., Cuillandre J.-C, 

Fort B., Mellier Y., 1998, A&A, 334, L9 
Miller, S. T. & Veilleux, S. 2003, ApJS, 148, 383 
Morrison H. L., Miller E. D., Harding P., Stinebring D. R., 

Boroson T. A., 1997, AJ, 113, 2061 
Pohlen M., Dettmar R.-J., Liitticke R., Aronica G., 2002, 

A&A, 392, 807 
Sackett P. D., Morrison H. L., Harding P., Boroson T. A., 

1994, Nature, 370, 441 
Schlegel D. J., Finkbeiner D. P., Davis M., 1998, ApJ, 500, 

525 

Wolf, C, et al. 2004, A&A submitted, astro-ph/0403666 
Zepf S. E., Liu M. C, Marleau F. R., Sackett P. D., Graham 

J. R., 2000, AJ, 119, 1701 
Zheng Z., et al., 1999, AJ, 117, 2757 
Zibetti S., White S. D. M., Brinkmann J., 2004, MNRAS, 

347, 556 (Z04)